I have been reading the methodological background and references used by Expostat and found the use of the 9/4 substitution method for the right censoring point. I would like more information about this procedure and what reference is used.
Welcome and thanks for posting. At first I did not understand your question, as our paper does not mention this approach to right-censorship. Right censored data in the inferential part of Expostats are treated by the bayesian model the same way as left and interval censored data are.
However it is indeed used in the “descriptive tab” of the tool. This tab provides descriptive statistics in order for users to check whether the data is reasonably lognormal and there was no particular issue for the tool to read it. Since the objective is just a basic description of the sample, and some statistics (mean, sd, GM, GSD) and graphs (QQplot) require a numerical value, we treated censored data as described on the website : ROS for left censored, middle point for internal censored and 9/4 censoring point for right censored. The latter factor is ad-hoc, and although I have seen it used, I never found any formal reference for it.
I wish to insist however, that the inferential part of expostats, were we calculate risk metrics as well as estimate parameters from the underlying exposure population (such as GM, GSD), does not rely on these simple methods, but on the bayesian treatment of censored data.
So for instance the GM and GSD values in the descriptive tab should be seen as very crude, and not be used for inference (we might well eliminate them from this part of the tool in the near future because of this confusion).
I hope this helps